[evla-sw-discuss] Keeping track of spectral windows at the EVLA: SDM vs. BDF

[Michael Rupen]

> Michael Rupen wrote:
>>  Hello all --
>>
>>  we've gone a full round on the enumerations and on the ordering of SpW,
>>  so here's the final decision (for now, anyhow :) --
>> 
>>
>>  1- On the enumeration front let's stick with what Sonja has just
>>  put in:
>>
>>  A1C1_3BIT A2C2_3BIT AC_8BIT B1D1_3BIT B2D2_3BIT BD_8BIT
>>
>>  I believe that order is alphabetic and deterministic.
>>  - Note that as Barry says we will never have, e.g., both A1C1 and AC,
>>  so the relative ordering of AC and A1C1 doesn't really matter.
>
> I'm preparing to implement this and want to point out something in case it 
> was overlooked. This is not really an issue for OSRO, but i'd like to code 
> for post-osro now if possible.
>
> The A0/C0 (or A/C, 8-bit) IF pair can lead to TWO widar baseband pairs: pair 
> 0/2 and pair 1/3. Each of these pairs would then have the same "AC_8BIT" 
> value for the new swbbName attribute.
>
> Is this OK?
>
> If so, does it mean the creation of the new swIndex attribute of the subband 
> needs to be created by sorting subbands across these two BBPs (0/2 & 1/3)?


To summarize our discussion of a few moments ago:

Yes, this is OK.

The combination of swbbName and swIndex must be unique.

swIndex for each swbbName must start at 1 and count up from there via
consecutive integers (1,2,3,4,...).

The algorithm for ordering the subbands (i.e., choosing which swIndex
goes with which subband) remains the same, with the addition that all
subbands for AC_8BIT are lumped together when determining their order,
regardless of their bbid's (0/2 or 1/3).

If we get to the last step (i.e., subbands identical except for sbId's),
subbands from baseband pair 0/2 should come before subbands from baseband
pair 1/3.

Eventually we may keep track of baseband ID etc. in the SDM, but not yet,
and possibly never.

               -- Michael